Learning algorithms have been implemented to eliminate the impulse distortions found in old records. The neural network simulation was tested with regard to the detection of hiss, scratches, and clicks. Various algorithms for the interpolation of distorted signal fragments were studied and implemented practically to the sound restoration. The quality of the restored sound patterns was assessed subjectively. The applied methods, results of subjective testing, and conclusions are presented in the paper.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.